Download Egocentric and Sociocentric Structure in

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Marriage wikipedia , lookup

Ethnicities of the Philippine Cordilleras wikipedia , lookup

Incest taboo wikipedia , lookup

Kinship wikipedia , lookup

Structural anthropology wikipedia , lookup

Western European marriage pattern wikipedia , lookup

Family wikipedia , lookup

Transcript
Egocentric and Sociocentric Structure in Classificatory Kinship Systems: Four Theorems
Douglas R. White
[email protected]
University of California, Irvine, Anthropology and Mathematical Behavioral Sciences
Abstract: The feature of Dravidian kinship terminology is typically that male lines on ego’s “side”
marry and call their “affines” relatives in a set of opposing male lines. The egocentric versus
sociocentric debate in Anthropology over the social network implications of Dravidian
terminology is resolved with proof of a single theorem: For a connected network A of marriages
between consanguineals, including only the additional ancestral relatives leading back to the
consanguineal ancestors of those couples, then if the kin of the couples are consistently sided
egocentrically, according to Dravidian kinship terminology, then all relatives in network A are
consistently sided sociocentrically, whether sides are defined through opposing sides V of male
kin, U of female kin, or both. Two other theorems prove that if all the consanguineal marriages
in network A are same generation (same number of generations back to the common ancestor for
the husband as for the wife) then if sidedness is V it is also U, if U it is also V. Finally if network
A is both U and V then all of its marriages are same generation and the marriage structure of A is
one of implicit alternate-generational moieties, as in a Kariera kinship network.
Classificatory kinship systems have been a vexing problem for social anthropology. Lewis Henry
Morgan, inventor of kinship studies and discoverer of “classificatory” versus “descriptive”
kinship, failed to understand that the South Asian “Dravidian” type of kinship system differed
radically from the more common use of kin terms in which father and father’s brother are
equated (similarly for mother and mother’s sister) and opposed to the opposite-sex sibling.
Bifurcate merging of this sort (as found for Iroquois terminology, for example) was
differentiated from a seventh type of kinship system beyond Morgan’s original six by Lounsbury
(1964), who mapped the features of the Dravidian system. Dravidian not only distinguishes kin
in the basis of cross-sex sibling links in the ascending generation, but uses bifurcation of the
male descent line of a brother and an opposing male line into which the sister marries. This is
called a “two-line” system (Dumont 1953), referring to egocentric kin terms, and implicitly
male-oriented. (No empirical cases have been found for the logical possibility of a “two-line”
system where the opposing lines are female.) Newer definitions of classificatory and descriptive
terms, however, are given by Dwight Read (2001, 2007, 2008, 2009, 2010a, 2010b, Read and
Behrens 1990), expert in generative mathematical models that match cognitive processes of
language use to linguistic structure in relative products denoting relations such as
(brother)°(father)=(uncle) where the output of composition is another term. His findings show
that Morgan was intuitively correct that Iroquois and South Asian (Dravidian) kinship
terminologies are both classificatory, although with important differences (Inset 1).
Read’s definitions (in Inset 1) are stated from an egocentric viewpoint: how does ego generate
kin terms as a composition of terms? I will look on these terminologies in the classificatory
sense, as usually defining exclusive categories into which one may marry. Dravidian, for
example, is associated with a rule of marriage to a classificatory “cross” relative, although not
exclusively. Australian systems of classificatory terminologies have an associated marriage rule
that one may marry only a person in the proper classificatory “cross-cousin marriage section”
that entails a sociocentric view of how actual marriages arrange themselves in relation to the
1
terms used for relatives. From a sociocentric perspective, Australia ethnologists sometimes use
“eight line” and “four line” for variants of the Aranda kinship systems, “two line” for the Kariera
kinship system and “one line” for the Aluridja kinship system (the latter with an age difference at
marriage so great that everyone marries into the same slanting “generation”). These refer to
“sections” with classificatory marriage rules that purportedly link actual behavior with kin term
categories and thereby define a kinship “system” (Denham and White 2005).
Inset 1: Dwight Read (comments edited by DRW)
Definition: Classificatory terminologies are those that are generated by beginning with both
an ascending generating term (e.g. 'parent', 'father' or 'mother') and a "horizontal"
(sibling) generating term. The merging criterion is a consequence of having both an
ascending and a sibling generating term. The most straightforward logic for generating a
structure of (male) ascending and descending kin terms leads to o/y same sex sibling
terms in most classificatory terminologies.
Definition: Descriptive terminologies are those that are generated by beginning with just an
ascending generating term.
In general, marriage behavior does not form a “system” with respect to kin term categories
except in a limited sense. Most societies observe an incest rule, for example, that is linked to kin
terms. One might be prohibited from marrying a sister, a first cousin, first to third cousin, certain
subtypes of cousins, and so forth. Similarly for marriage with an ancestor or a descendant. The
restrictive classificatory marriage rules of the Dravidian or Australian systems, however, might
be called “elementary,” to rephrase Lévi-Strauss (1949), while Iroquois would be better
designated as “semi-elementary” because of fewer marriage restrictions, while societies with
only incest prohibitions would fit Lévi-Strauss’s kinship designation of “complex systems.”
Lévi-Strauss (1949) regards as “semi-complex” systems that prohibit marriage so broadly as to
become restrictive, as with the Crow-Omaha types of kinship terminology.
The features of classificatory and descriptive kinship listed in Table 1 do not entail a linear
evolutionary sequence. Two alternatives for branching proto-evolution might be: (1) from protoDravidian (often considered in recent literature an archaic kinship type) to each of the other
classificatory types is plausible; (2) from proto-Iroquois (the simplest system) in alternate directions
for expanding kin term networks is also plausible – either through extending classificatory terms, or
increasing local restrictions to extending marriageable kin at greater distances (Iroquois, CrowOmaha).
Classificatory systems Older/younger distinctions
spouse kin term
marriages
sibling,
symmetric
cross-cousin, sibling, symmetric
wife=cross-cousin
wife≠relative
2 2-line m,f
2 line male
sibling (same sex, opposite)
sibling same sex, asymmetric
wife≠relative
wife≠relative
ranking?
“Elementary”
Kariera
Dravidian
“Semi-Elementary”
Iroquois
Oceanic/Polynesian
“Semi-Complex”
Table 1: Features of Classificatory kinship
2
Mathematical logic and nonmathematical argument. Mathematical inference, however, is
stronger than typologies and labels: it offers proofs of equivalence or implications. It may also
provide a means to resolve lingering disputes about kinship “systems.” The disputes that I
resolve here with proofs of mathematical theorems are between views of Dravidian
terminologies from “egocentric” and the “sociocentric” perspectives, which differ in the
Oceanic/Polynesian, Iroquois, Australian and Dravidian classificatory logics.
Sidedness: Egocentric and Sociocentric
Egocentric sidedness: Definitions. Individual consanguineal marriages may be viri-sided and/or
uxori-sided or neither, as defined by Houseman and White (1998a). For spouses with a common
ancestral couple, counting male and female links to that ancestor (including the husband and wife
themselves as links) let F be the sum of female links, F = fm + fw for the manm and womanw
respectively. A viri-sided marriage is one where F = even. A uxori-sided marriage is one where G =
gm + gw = even, with g and G terms being male links for the manm and womanw respectively. Thus
for FZ/BS G = 2 = even and F = 1 – odd, which is uxori-sided and not viri-sided. F = G = odd is a
marriage that is neither uxori- nor viri- sided (e.g., FFZDD, F = G = 3 = odd). F = G = even is both
uxori- and viri- sided (e.g., MBD and FZD, F = G = 2). Figure 1 (left) shows a viri-sided marriage
with G = odd heavy lines for sons and F = even light lines for daughters. The first and third line of
males form side 1 and the middle two classificatory lines of males form side 2. Figure 1 (right) shows
a uxori-sided marriage, where G = even. It is not required that there be as many lighter lines
ascending generations as descending from the common ancestral couple.
Figure 1. A viri-sided marriage, F = even
G=3, male; F=4, female links to
a common ancestral couple
1
2
Uxori-sided, G = even
F=3; G=4
(gender lines and
sidedness reversed)
3, where line 3 folds to → side 1
Sociocentric sidedness: Definitions. A connected consanguineal marriage network or CCMN (with
“consanguineal endogamy“) is a kinship group or subgroup CCMNS in which each marriage is with
a consanguineal relative. The network includes the common ancestral couple and those who link to
them. The network is viri-sided (uxori-sided) to the extent that its individual marriages are viri-sided
(uxori-sided), as in Figure 1 (the right graph reversing the genders of the left). Sociocentric
sidedness, however, requires that the egocentric sidedness of individual marriages are aligned so that
within the consanguineal marriage network, if s(x) is the set of same-sided relatives of x and o(x) are
the opposite-sided relatives of x, then s(x)
s(s(x)), s(x)
o(o(x)), o(x)
s(o(x)), and s(x)
o(o(x)), i.e., sidedness is associative. Sidedness distinctions overlap with but differ from those of
cross/parallel relatives.
Kinship Terminologies. Read details the differences among classificatory kinship terminologies as
follows (Inset 2):
3
Inset 2: Dwight Read: (comments edited by DRW)
All classificatory terminologies are generated with an ascending and a
sibling generator that combine to generate bifurcate merging and, with few
exceptions, to generate o/y same sex sibling terms (more accurately,
"ascending", "descending" sibling distinctions).
Dravidian terminologies are distinguished from the Iroquois not by the
generators, but a further step in the generation of a terminology in which
a structure of male terms and structure of female terms are linked to form
a single structure. This accounts for the symmetry in o/y cross-cousin and
sibling terms. Like Iroquois, Dravidian lacks the terminological equation
of spouse = 'cross cousin.' A "marriage rule" that does not require samegeneration marriage is emergent through a "two line" male-sided marriage
opposition in the middle three generations (as discussed by Dumont, who
does not translate marriage opposition into sociocentric sides)*. The
terminology constrains marriages to 0 generation with cross-cousin, -1
generation marriages with "ZD", and +1 generation marriages with
classificatory "MB" ("FZ" in a female-sided "two line" marriage opposition
is also possible but is highly unlikely if only for demographic reasons).
Kariera, and presumably other Australian terminologies, have two logics:**
(1) The generating logic equates, spouse = 'cross-cousin' and child of
'cross-cousin' as 'child' as part of the terminology and thus a crosscousin marriage rule and section system, reversing the logic of Iroquois.
(2) A sociocentric logic of section systems as an intersection between a
“two line” male-sided marriage opposition cross-cut by a "two line" femalesided marriage opposition, thus closely related to Dravidian.*** In either
case the structure has the section system as an emergent sociocentric
property and accounts for the symmetry in the o/y sibling terms.
The Iroquois terminology has o/y sibling terms and is generated (in
contrast to Kariera) with the child term for male and female cross-cousin
terms distinct from own child. (This obviates the cross-cousin marriage
rule in the Kariera terminology and eliminates an emergent section system
structure). It entails that the Iroquois terminology has no cross-cousin
marriage rule, and no sociocentric sides.
The Polynesian and Oceanic terminologies and kin terminology lack a cross
cousin marriage rule as part of the terminology structure. (In fact they
typically do not have cross cousin marriage or marriage rules.) This also
accounts for the asymmetry of o/y sibling terms (only for same sex
siblings, but not for opposite sex siblings) in these terminologies.
The generative logic of these systems accounts for precisely the properties
of the terminologies, the presence/absence of a marriage rule framed or not
around cross-cousin, the pattern of o/y sibling terms and, perhaps most
important, has ethnographic support for the different ways in which the
structures of male and female terms are joined through ethnographic
observations that show that the logic is realized concretely in conceptualizations about siblings in societies with these kinds of terminologies.
*Conditions for this are proven by White, below.
** (1) Read’s view, (2) White’s view: proven below to be convergent, the discrepancy disappears.
*** The proofs herein show conditions under which any two of the two forms of sidedness and same
generation marriage entail the third.
4
Sociocentric sidedness: Theorem 1. A consanguineally endogamous CCMNS with F = even or G =
even or both F = G = even for all its consanguineal marriages will be sociocentrically, in turn: virisided, uxori-sided or both.
Proof of Theorem 1.
If F = even then for a consanguineal marriage the Hu’s side (parallel kin) will include his patriancestors (PAs) and there are S PA groups including and from the Wi’s PA to new PAs through
other maternal links, terminating in Hu and Wi’s common ancestral couple, as in Figure 1 (S = 2),
and then S’ PA groups from the ancestral couple through daughter and son links back to and
including the Hu’s PA (S’ = 2 in Figure 1). In Figure 1 S + S’ = 2 + 2 = even, so every even
numbered PA in the cycle of PAs from Hu’s to Wi’s common ancestral couple back to Hu’s PA
can be folded into side 1, and every odd numbered PA in the cycle can be folded into side 2.1 Thus,
every sided consanguineal marriage folds into two sides. Because any consanguineal marriage
folds into two sides, any two such marriages having a person C in common will fold into two sides:
C’s side, and the opposing side. In a connected consanguineal marriage network, for each
consanguineally married couple, the PAs in the entire network, all of which are by definition
connected will fold into two sides. The same proof follows if G = even or F = G = even. Q.E.D.
That is, egocentric sidedness will produce a consistent sociocentric sidedness structure where
consanguinity is the basis of reckoning sidedness. Figure 9-2 in Houseman and White (1998a),
including the consanguineally endogamous CCMNS for the Makuna, is a perfect example for F = G
= even for the consanguineal marriages (100% viri- and uxori-sided), i.e., F = G = even for all these
marriages, which also implies they are all same-generation (see Theorem 2). For all Makuna
marriages F = even with a single (1%) nonconsanguineal marriage exception. Houseman and White
(1998a) found “sided” kinship networks similar to the Dravidian (i.e., Dravidianate) in Amazonia,
and created percentage measures for the extent to which they were sociocentrically viri-sided, uxorisided, or both.
Theorem 1 is merely saying that since all connected consanguineal marriages in a kinship network
must be sided or unsided, then if they are all consistently sided (F = even or G = even or both F = G
= even) egocentrically, they will be consistently sided sociocentrically. Statistically, however, if we
count the number of independent genealogical cycles of common ancestries for the consanguineal
marriages, we would want the numbers of those that are consistently sided to be significantly greater
than the expectation that in random cycles, half will be sided and will not. If Theorem 1 does not
extend to nonconsanguineal marriages it is not because there are pairs of marriages connected in
cycles with two common ancestral couples where F = even or G = even does not imply sidedness. It
is because sidedness in these more complex nonconsanguineal marriages is not easy to compute.
Practical decision making. There is a practical explanation for the existence of the F, G types of
sidedness rules among Dravidians, dependent on consanguineal marriages for reasons of
computation. It is not that consanguineal marriages have a more powerful force of norms. Dravidians
in general, and Pul Eliya in particular, do not consciously attempt to construct sociocentric sides to
extend the coherence of their egocentrically sided kinship terminologies. Nor are sociocentric sides
“governed” by adherence to social norms. The Sidedness Theorem 1 proves that when individuals
are consistent in their egocentric behavior – the type of consanguineal marriages shown in Figure 1 –
1
Because F = even requires S + S’ = J = even then if S = even, S’ = even links through daughters and their PAs
back to Hu’s PA. If S = odd then S’ = odd. For every j=1,…, J the j = even PA’s are on the Hu’s side and the j =
odd PA’s are on the Wi’s side.
5
sociocentric consequences are automatic and inevitable even if unintended. The theorem is a
demonstration that anthropologists cannot afford to ignore mathematics, which can define objects
that have necessary connections between parts and wholes. This insight is particularly appropriate in
the case of networks (Chapter 1 of White and Johansen 2005 dwells on this at length).
There is, however, an intentional component to sidedness in consanguineal marriages, which is the
generator of sociocentric sidedness. It arises under conditions of uncertainty, when the parents of a
potential bride and groom (i.e., those who usually arrange marriage) have no preexisting kin terms
for one another, and so are uncertain about whether their children are marriageable. As is attested by
various South Asian ethnographers (in personal communications from Lehman 2010, Leaf 2010, and
Kolenda 2010), the practical expedient that is observed in conversations is to find a common
ancestral couple and count their numbers of F = {fm, fw} of female(-cross) links to ancestral couple.
If F = even then they are properly viri-sided and eligible to marry.
This viri-sided Dravidian marriageability decision rule, or one comparable, does not extend to
nonconsanguineal marriages. Mathematics provides the clue as to why this is true. Our hypothetical
pair of parents arranging a marriage need only search among their two sets of parents (matching in a
space of 22), their two sets of grandparental couples (matching in a space of 4 2), or their two sets of
great-grandparental couples (matching in a space of 82), a well defined problem. When there are no
common ancestors, they could consider whether there are cycles of marital relinking (like sister
exchange or marriages of two brothers to two sisters, where marriage between offspring is proper),
but beyond that short range of affinal kin, there are too many possibilities and in no particular order
to formulate a consistent and effective decision rule.
Harary (1953) proved a balance theorem of signed graphs – and using Figure 1 as an example of a
marriage graph – which establishes, if we label the heavy lines in as positive (+) and the light lines as
negative (-1), that if the product of signs in every such cycle is positive, then the signed network
consists of two sides, each connected internally by positive links and externally by negative links
between sides. This is easy to conceptualize, prove, and compute with appropriate software, but is
hardly something that can be implemented as a decision rule. Its import is that if sidedness (as an
instance of a balance principle) is implemented locally then the network structure is globally sided,
and the reverse: global sidedness entails that every cycle is balanced in terms of an even number of
crossovers between two sides.
The practical rule of Dravidian sidedness computation according to common ancestral couples is thus
a practical and effective solution to a complex problem. It leaves uncertain whether a potential bride
and groom who have no common ancestors are marriageable. This is often the situation when
outsiders marry into a local Dravidian community. Here sociocentric sidedness will be lacking,
which is exactly what we find in the Pul Eliyan marriage networks. All Dravidians have a term for
wrong or improperly sided marriages and also a term for proper marriages (e.g., murai in Nattathi
Nadars Tamil). Two brothers from a remote village might marry women from opposite sides of the
Pul Eliyan village sidedness structure, and these might be recognized as “wrong” marriages. Pul
Eliyans might choose to ignore the brothers’ same-sidedness stemming from their home village,
however, since it has no practical effect in terms of local sidedness, where in-laws are important in
labor and property exchanges.
Kariera and Dravidian: Definitions relevant to proofs of further Theorems
P-graph. Following Weil (1949), but for concrete marriages rather than marriage-types, "let the
number of types of marriages be n", and let each marriage be numbered from 1-n, as in Figure 1 (left,
6
n = 7 marriages with 14 spouses), where ♂ links (agnatic descent lines) are denoted by heavier lines
and ♀ links by lighter lines. A P-graph has vertices that represent unmarried children or couples
(with or without children) and P-links between vertices directed from child-to-parent(s). The inverse
links, parent(s) to offspring, are given by p=P-1. Vertices represent single individuals in the case of
unmarried children, whose type of P link (♂male, ♀female) depends on their gender. All other
vertices represent either a single parent (partner unknown) or a parental couple. Their P links to
parents again depend on gender: husband to parents, wife to parents. A sub-P-graph Q consists of a
set Q of vertices and all the P links between vertices in QxQ. Although one may have more than one
marriage, P-graphs normally represent (1) real or fictive genealogy wherein no-one is their own
ancestor, (2) two-parent opposite-sex marriages, (3) only one set of parents for a child. Thus: (4) no
vertex will have more than two parental links, one to a male’s parent(s) (for a couple: husband's), the
other to a female’s (for a couple: wife's) parent(s). Other variants of P-graphs may be defined.
P-graph generations. The P-graph defined by 1-4 above is a directed asymmetric graph (DAG) with
1 ≤ gen ≤ g generations, g equal to the length of the longest directed path (e.g., g = 4 in the Figure 1
marriages), which insures that every generation will have at least one parent in the preceding
generation. Program Pajek (Batagelj and Mrvar 2008) computes g generations for P-graphs.
Sides and sidedness. These are implicit moiety structures in the marriage network, wherein a P-graph
can be uniquely partitioned into two, within each the vertices are connected by the links of one
gender and between which they are uniquely connected by the links of other gender.
V= Viri-sides. Here, a P-graph or sub-P-graph Q is split into two sides, uniquely determined, within
which vertices are connected by ♂son/parent links and between which they are uniquely connected
by ♀daughter/parent links.
U= Uxori-sides. This a P-graph or sub-P-graph Q that is split into two sides, uniquely determined,
within which vertices are connected by ♀daughter/parent links and between which they are uniquely
connected by ♂son/parent links.
S= Same-generation marriage. In general, siblings do not all have to marry spouses of the same
generation, nor does “same generation marriage” imply that generations partition by time periods,
nor that men and women share the same temporality of generations (when women marry on average
at a younger age than males, their generational time is faster than males, for example). Nonetheless, a
P-graph component (connected graph) connected only by P°p (sibling or parent’s child) links among
marriages constitutes a pure generation if and only if it contains no connecting parents P-connected
to other connecting parents within it (i.e., no grandparents). If the pure generations of the P-graph are
partially ordered by the g generations in the P-graph, then the P-graph has property S, same
generation marriage. A sub-P-graph Q may also have property S.
A= Alternate generational moieties are implicit when a sub-P-graph Q is same-generation but a
larger subset R  Q has some marriages with both same-generation marriages and ones between
individuals +2 and -2 generations apart, reciprocally but none at +1 and -1 generations apart.
B= Bicomponent. A bicomponent of a marriage network P-graph containing marriages x and y is a
maximal sub-P-graph B containing x and y in which for any node z there are two or more disjoint
paths between very pair of nodes in B. (All links between nodes in B that occur in the main graph are
by definition also in the P-graph)
7
Ck= A sub-P-graph Q is one of consanguineal marriages if and only if every marriage within Q for
which there are two parent links is between a male and female who have a common consanguineal
ancestor less than k generations back.
Sociocentric sidedness: Theorems 2-4.
The presence of both viri-sides (condition V) and uxori-sides (condition U), for a sub-P-graph
marriage network Q of Ck consanguineal marriages, in a bicomponent B of a marriage network,
logically entails alternate generational moieties (conditions S and A) in B. This includes the
possibility of an ego at generation i, 1 ≤ i ≤ g, marrying someone at generation j, 1 ≤ j ≤ g, where the
absolute difference |i - j| is an even number, e.g., +2 or -2 generations (condition A). Further, the
presence of S, same-generation marriage, and either viri-sides V or uxori-sides U logically entails the
complementary type of sidedness.
Theorem 2. U, V, and Ck, consanguineal marriages => S and A: implicit alternate-generational
moieties.
Theorem 3. V and S => U
Theorem 4. U and S => V
Proof of Theorem 2.
Suppose U and V and consanguineal marriages. For a sub-P-graph Q in which every marriage is
between consanguineals (Ck, consanguineal marriages), then marriages in which all husband and
wife marriage cycles are with ancestors are both U and V, and each such cycle will have even
number of female links, an even number of male links, and adding the two, an even number of total
links. Thus, if the ancestral graph for each such cycle is drawn, from parent to child in adjacent
generations, then either husband and wife will be of the same generation or one is an even number
of generations above the other. This will apply to all such marriages, and conditions S and A will
be satisfied: implicit alternate-generational moieties. Q.E.D. (This proof generalizes to Australian
section systems.)
Counterexample viri-sided and uxori-sided nonconsanguineal marriage lacking the feature of
same-generation marriage
Suppose U and V and a network without Ck, consanguineal marriages, in which a man marries a
BDHBD (as in Figure 2), implying two nonconsanguineal marriages (they are linked by two
distinct ancestral couples, not a common ancestral couple). Then the number of males is even (4) as
is the number of females (2), qualifying for U and V but S and A are violated.
Figure 2: BDHBD as an example of a nonconsanguineal viri-sided and uxori-sided marriage
(F = 2, G = 4) violating same-generation marriage for siblings
Lemma 1.
Suppose A (Implicit alternate generational moieties). This may apply so that the generation number
of each marriage is the required minimum of one below the generation number of the parents of the
8
husband and that of the wife, the other alternative being an odd number below. This entails that any
marriage cycle will have an even number of links because for any distance d (odd) from one spouse
to the common ancestral couple, the distance is a d’ (odd) for the other spouse, for a sum of two
odd numbers, which is an even number of total links in every such cycle.
Proof of Theorem 3.
Suppose V and A. By Lemma 1, A requires that every marriage cycle will have an even number of
links ec, while U entails an even number of male links em, so the remaining female links ef = ec - em
must be even, and the network is U. Q.E.D.
Proof of Theorem 4.
Suppose U and A. Exchanging U and V above: then V and A entail V. Q.E.D.
Evidence that Dravidian is viri-sided and not uxori-sided. (1) Ethnographers report that the
Dravidian marriageability decision rule is viri-sided and not uxori-sided. (2) Trautmann (1981)
reports that MB/ZD marriages, which are viri-sided and not uxori-sided, occur in Telugu and Tamil
(core Dravidian) systems, and White (1999) and Houseman and White (1998b) show extensive
classificatory ZD marriage in the connected consanguineal marriage networks (consanguineally
endogamous CCMN) of Pul Eliya. (3) All of the Dravidian kinterm systems reported in Trautmann
(1981: 121, 134, 135, 138, 141, 144, 150:Tamil, 154: Sinhalese, 156, 157, 159, 162, 163, 165,
166:Telugu, 170, 188) show ZD as “cross” (consistent with viri-sidedness) while BS (consistent with
uxori-sided marriage) is “parallel” and thus unmarriageable. All the evidence from triangulated
sources (actual marriages, kin terms, marriageability decision rules) points to Dravidian as viri-sided
in terminology, actual marriages possibility, and in decisions rules – and contradicts the possibility
that any if the Dravidian kinship systems are uxori-sided or that they necessarily prescribe samegeneration marriage only.
A Sinhalese Example of the Dravidian Kinterm Systems
Figure 3 illustrates one of the core Dravidian kinterm systems, that of Sinhalese, with notes on the
male terminology and kin behavior given by Leach (1961:126) and the Tamil terminology studied by
Read (2010b). It will be noted that the MB/ZD reciprocal terms are both cross and allow marriage
while the BS/FZ terms disallow marriage because BS is parallel. Here, as in Dravidian generally,
wife=”woman” and there is no further designation of a kin term for wife or husband (also confirmed
by Kolenda 2010, along with Tamil ZD marriage). Hence cross-cousin marriage is not prescribed and
there is flexibility in the possibility of classificatory MB/ZD G+/-1 marriages that do not disrupt prior
kin term usage.2 A classificatory MB marrying a ZD, however, will call her mother X- “aunt” after
marriage (“FZ”) as if the wife’s mother were the “FZ” and not the “Z,” a convenient omission for a
viri-sided system. Trautmann (1981:206-207), in salvaging his “cross-cousin only” marriage rule
assertion for Dravidian, brushes away all ZD marriages in Dravidian societies as if they were
Brahmin exceptions or as if ZD marriage was an anticipation of a FZD marriage. Marriageable
categories are shown in heavy outline in Figure 3. Unlike Kariera, with four G+/-2 distinctions
reflecting viri-sided and uxori-sided section systems, there are only two G+/-2 distinctions that do not
neutralize the existence of viri-sidedness in the sociocentric structure of viri-sided marriages in the
consanguineal network but do neutralize the possibility of G +/-2 marriages. The e/y (o/y) relative age
distinctions for same sex siblings, typical of classificatory kinship, are present and are extended to
2
The lack of kin terms for husband’s relatives also facilitates the retention of prior kin terms for consanguineal
relatives of different generations who marry, such as MB/ZS. Thus, after marriage, the husband is still MB.
9
the parental generation.3
Figure 3. Paradigm of Sinhalese kin terms (Trautmann 1981:154; Figure 3.20) and Pul Eliya.
Reciprocal kinship behaviors between males (Leach 1961:126) are described in the footnotes.
♂
Sinhalese
♀
X cross
// parallel
X cross
2
4
G
kiriāttā
kiriamma
FF, MF
FM, MM
1
6
7
G e
loku appā FeB, appā
amma
loku MeZ0,
māmā 5
MeZH
FeBW0
nändā 8
y MB,FZH, WF bāppa 3 FyB,
F
M0
kuda MyZ0,
FZ1,MBW1,
MyZH
FyBW0
WM1
0
10
G e
ayiyā eB, e(FBS),
akkā eZ0, e(FBD0),
massinā 9
e(MZS),e(MBDH),
e(MZD0),e(MBSW0),
nänā 11
e(FZDH)
e(FZSW0)
y
MBS, FZS,
malli 6 yB, y(FBS),
namgi yZ, y(FBD),
MBD3, FZD3,
WB, ZH
y(MZS), y(MBDH),
y(MZD),y(MBSW),
WZ3, BW 3
y(FZDH)
y(FZSW)
G-1*
bänā 2
putā 3
duvā
leli 2, 13
12
9
9
♂?ZS, DH
S, ♂?BS
D0, ♂?BD0
♂?ZD4, SW4 9
G-2
munburā (miniburā) 1
minbiri
SS, DS
SD, DD
3
The question marks (?) before kin terms at G-1 could be removed if we rely on consistent patterns from other
Dravidian terminologies as to distinctions among these terms.
4
Kiriāttā / miniburā: Friendly informality. G+/-2. Divisions (♂ and ♀) apply to ego’s and spouse’s grandparents in
Tamil. All underlined kin terms are general Sinhalese but also hold for Pul Eliya.
5
Māmā / bänā: Respect but much less than between father/son (extreme when son-in-law is binna-married). G+/-1.
Divisions (♂ and ♀) apply to ego’s and spouse’s parents in Tamil. The reciprocal ♂Māmā / ♀leli classificatory
category for Pul Eliya contains distant consanguineal marriages that are properly viri-sided.
6
Loku appā or bāppa / putā: Respect relationship rather lacking in feeling on both sides. G +/-1. Divisions (♂ and ♀)
apply to ego’s and spouse’s parents’ siblings in Tamil.
7
Appā / putā: Extreme respect tending to avoidance. G+/-2. Divisions (♂ and ♀) apply to ego’s and spouse’s parents’
siblings in Tamil.
8
The opposite-sex reciprocal of FZ (X cross) is ♂BS=puta (//=parallel) hence unmarriageable, unlike MB. No term
for HM is attested by Trautmann, who uses seven different sources, including Leach (1961:126), which allows G+1
consanguineal correctly-sided marriages to be contracted without a conflict in egocentric kin terms.
9
Massinā / massinā: Familiarity tending to joking relation. G0. Divisions (♂ and ♀) apply to ego’s and spouse’s
cross-cousins in Tamil.
10
Ayiyā / malli: Marked respect, formality. G0. Divisions (♂ and ♀) apply to ego’s and spouse’s siblings and
parallel-cousins in Tamil.
11
There is no determination of the position of wife. Nänā≠wife. In Dravidian generally, wife=”woman”, as in: Baiga
(dauki=wife, woman), Vedda (gani= wife, woman), Kondh (ayal= wife, woman). In Nakarattar, a merchant banking
class we find descriptive term wife=pentir. There are many colloquial terms and ways of referring to wife, however,
for example the Sinhala terms mahattaya (“husband”) and nona (“wife”). These are actually status terms such as
(doctor sir/lady) conveying the sense of not only husband and wife, but also master and mistress. G 0.
12
Question marks for G-1 are imputed by consistency with other Dravidian systems (Trautmann 1981:40, 103, 121,
134, 135, 138, 141, 144).
13
It is the ♂eZD that is eligible for marriage with ♀MyB in Tamil and Karnataka but not in Sinhala/Pul Eliya.
10
Trautmann’s (1981) characterization of Dravidian terminologies as accompanied by a “same
generation” marriage rule and a “classificatory cross-cousin” marriage rules is clearly incorrect. It
follows that the Tjon Sie Fat and Trautmann (1998) formal analysis of Dravidian terminologies is
also wrong as it follows the “same generation” marriage rule and “classificatory cross-cousin”
marriage rule assumptions of Trautmann (1981). In turn, the Barbosa de Alameida (2010)
formalization of Tjon Sie Fat and Trautmann’s (1998) Dravidian semantic logic is wrong. White
(2010) addresses these issues and offers repairs. These repairs are needed for the viewpoints on
Dravidian terminologies made by Godelier, Trautmann and Tjon Sie Fat (1998) in their edited book.
Conclusions
The intent of this careful ethnological survey and the mathematical theorems that help explain some
deep anthropological issues is to convince anthropologists and historians of the evolution of kinship
systems that proofs of equivalences or convergence between different concepts that have an
empirical foundation are useful in understanding cultural phenomena and the links between kinship
terminologies, cognitive processes that are culturally shared, decision rules, and marriage behavior.
Theorem 1 is a proof of something long denied in the social and historical anthropology of South
India, and applying equally elsewhere: the egocentric sidedness of classificatory Dravidian kinship
terminologies, when put into consistent practice among consanguineally related kin, is strictly
equivalent to sociocentric sidedness whereby connected consanguineal marriage networks (which
form the core of many Dravidian communities) divide into an empirical “two line” system, with male
lines between which females marry, i.e., between sides.
Houseman and White (1998a) define this type of “two line” system as viri-sided, and distinguish it
from uxori-sided “two line” systems. No genuine case of uxori-sided “two line” kinship systems have
yet been discovered, although there do exist many cases of named matrimonial moieties in which
moiety membership is transmitted matrilineally. The Dravidian viri-sides are typically unnamed, and
are not associated with patrilineages but rather with cognatic descent. In the case of Pul Eliya,
females may inherit what is normally agnatic property (compounds and irrigation rights) when
brothers are lacking.
Houseman and White (1998a) and White (1999) show that the special case of consanguineal
marriage network sociocentric viri-sidedness in Pul Eliyan is 100% consistent with Leach’s (1961)
marriage data. Outside of this community core of consanguineal marriages with common ancestral
couples, viri-sidedness is very commonly violated. Most of these violations are for marriages with
outsiders or members of specialized occupational castes living within the community. The female
inheritors of agnatic lands, however, whose marriages could violate the connection between agnatic
landholding groups that are consistently “sided” across generations, adopted in every case the
sidedness of their missing brothers and avoided making “wrong marriages” by marrying men from
distant villages whose “sidedness” with respect to Pul Eliya need not be recognized. These marriages
are among those that Leach noted were uxorilocal-residence “binna” marriages (sisters retaining the
agnatic compound) as opposed to the “diga” virilocal-residence marriages normally practiced by sons
of compound-owning families. Given the considerations of cognatic inheritance, then, it is obvious
why the “sides” of Pul Eliyan kinship were neither formally named nor inherited patrilineally as
named moieties.
The theorems proven here about sidedness in connected consanguineal marriage networks
(consanguineally endogamous CCMN) by no means imply that larger marriages networks with
egocentric sidedness consistent with Dravidian terminology will be sociocentrically sided. Figure 4
11
shows an example of how the theorems apply for the case of Pul Eliya (Houseman and White
1998b). In this figure male lines connect identically colored marriage nodes that are identified with
Leach’s (1961: flyleaf genealogy) compound number identifiers for husband & wife in each
marriage. At the lower right, for example, couple H3&C1:4 refers to a Compound H:3 grandson of
H:1 married a Compound C1 great-granddaughter. Dotted arrows link daughters in a lower
generation marriage to her parents; solid lines do so for sons to their parents in a higher generation.
Parallel red lines emphasize two MBD marriages. Triangular red lines emphasize two FZD
marriages. These marriages form a consistently two-sided sociocentric pattern consistent with
Theorem 1. The solid oval encloses a set of marriages that are not consistently sided but they do not
form a connected consanguineal marriage network (consanguineally endogamous CCMN) and are
not expected, by Theorem 1, to be sociocentrically sided. The non-sided male line that is colored red,
for example, has ancestor v129 from an unknown outside village. A number of wrong-sided
marriages are discussed by Leach but not are part of the CCMN.
Figure 4. A p-graph of the Pul Eliya genealogy with generations steeply sloping to the upper right
and male descent lines sloping gradually to the bottom right). It shows three MBD marriages (red and
double black parallelograms) and two FZD marriage (a two–generation same-side male line in black,
and a two–generation between-side female line in red). Off-generation G+/-1 marriages are not shown
but are present in the network when more marriages are included. The center oval shows families
whose marriages mostly violate sidedness but are not part of the consanguineal marriage core of Pul
Eliya. Consanguineal marriages occur in both early and later generations: One is a MBD marriage
within Pul Eliya, and two between a Pul Eliyan and a FZD (in one case also a MBD) from another
village. Marriages of Pul Eliyans with a MBD of another village provide land claims in both villages.
There is, then, a need for social anthropology to recognize the existence of moiety-like marriage
structures that are unnamed, distinct from descent groups, and only partially implemented within a
community. The Pul Eliyan ethnography of Leach (1961) was the basis for a successful critique and
12
deconstruction of the fundamental concepts of the Descent Group theory of British Social
Anthropology (Dumont 1971) but the implications of Leach’s work were never fully understood until
network analysis of actual genealogical and marriage data was implemented by White and Jorion
(1992) and Houseman and White (1998a, 1998b). This approach is now widely understood14 but has
lacked the formal mathematical proofs supplied here for South Asian Dravidian and that also apply to
Dravidianate kinship systems elsewhere.
Theorems 2-4 are proofs of concepts long discussed by anthropologists but brought to the fore by
Dumont (1983), wherein he brings to bear all of his essays on Dravidian kinship systems and his
comparisons with Kariera and other Australian systems of kinship. Kariera is often called a two-line
system, referring to male lines or “sides,” in contrast to three line (Ambrym), four line (Aranda), and
eight line systems (an Aranda variant, or hypothetical Murngin). The common “two line” feature of
Dravidian and Kariera has been the basis for some theorists to claim an identity between them. This
is disputed by Dumont (1983:200), who sees Kariera section system as clearly “built on two
complementary oppositions operating crosswise: (1) between two kinds of local groups, ideally
affines to each other [the “two sides” of male lines]; (2) between two kinds of generations which
bisect each local group and which, as particularized in each kind of local group, are linked one to one
by intermarriage” (i.e., intermarriage between local groups within generations). His logic, as opposed
to other hypotheses, is that “More probably, these characteristics [generational moieties, widespread
in Australia] are aspects of a universal tendency to group together alternating generations, a tendency
which would have found its perfect development in Australia.”
Theorems 2-4 offer a simpler explanation, one which again applies only to the case of connected
consanguineal marriage networks (consanguineally endogamous CCMN). Theorem 2 says that in the
case of CCM networks, the intersection of viri-sided (V) and uxori-sided (U) networks entails S,
same generation marriages, hence alternate-generation moieties. Theorem 3 identifies Trautmann’s
(1981) error of assuming that Dravidian terminological systems are consistent, at least in a
historically reconstructed proto- Dravidian, with same-generation marriage. The theorem proves that
if S is true and V is true sociocentrically (which Theorem 1 shows is true in CCM networks that have
egocentrically viri-sided terminology), then U is also true, so that Trautmann’s (pseudo-) Dravidian
would consist of an intersection of V viri-sides and U uxori-sides, along with alternating generations.
Trautmann’s schema for Dravidian, then, would entail a section system, one that is structurally
equivalent to the Kariera system, with the difference that the four types of grandparent terms that are
present in Kariera kin terms are latent in Dravidian. There is ample evidence that Dravidian systems
are viri-sided and not uxori-sided, which makes them distinct in three ways from Kariera: (1) kinship
terminology, e.g., Dravidian with two grandparent/grandchild terms, Kariera with four, (2) actual
marriages, both allowing classificatory cross-cousin marriages, but only Dravidian allowing G+/-1
marriages, and (3) marriageability decision rules, which prescribe classificatory cross-cousins only
for Kariera, but viri-sided marriage in G01 and G+/-1 for Dravidian.
These results push well beyond Dumont’s notions of sociocentric structure to actual sociocentric
network divisions, but demolish many of Dumont’s (1983) specific conjectures.
14
See http://kinsource.net/kinsrc/bin/view/KinSources/, where scores of genealogical datasets prepared for purposes
of network analysis have been posted.
13
Acknowledgements
I am deeply indebted to Dwight Read for maintaining with me a two-month intermittent stream of
emails concerning Dravidian and other terminological systems which he has studied in detail and
from which we have both benefitted. Our views have converged about the need to update the work
by Trautmann and other scholars on the Dravidian kinship terminology and how marriages are
organized within it. Hopefully my findings have been as useful to him as to me. Any errors in my
rephrasing of the Read’s kinship terminology principles are my responsibility. I thank Barbosa de
Almeida for agreeing to share his algebra of Dravidian kinship, building on Trautmann’s work, for
commentary in a special issue edited by Paul Ballonoff for the journal Mathematical Anthropology
and Cultural Theory, where this article was submitted. I also thank Kris Lehman, Murray Leaf, and
Pauline Kolenda, South Asian ethnographers, for answering my questions on Telugu and Tamil
marriage decision rules. My apologies to Trautmann, Tjon Sie Fat, and Barbosa de Almeida for my
critiques of what I see as deficiencies in the foundations of their work. Any errors of fact or inference
remain my own. As for the mathematical proofs, which typically receive careful inspection in peer
review, I will be very happy to receive corrections by others or publications of simpler proofs.
References
Barbosa de Almeida, M. W. 2010. On the Structure of Dravidian Relationship Systems,
Mathematical Anthropology and Cultural Theory. In press.
Batagelj, Vladimir, and Andrej Mrvar. 2008. Analysis of Kinship Relations with Pajek, Social
Science Computer Review, Vol. 26, No. 2, 224-246.
Denham, Woodrow W., and Douglas R. White. 2005. Multiple Measures of Alyawarra Kinship.
Field Methods 17(1):70-101.
Dumont, L. 1953. The Dravidian kinship terminology as an expression of marriage. Man 54: 34-39.
Dumont, L. 1983. Affinity as a Value: Marriage Alliance in South India, with Comparative Essays on
Australian. Chicago, IL: University of Chicago Press.
Dumont, Louis. 1971 (translated and edited by Robert Parkin). An Introduction to Two Theories of
Social Anthropology: Descent Groups and Marriage Alliance. New York: Berghahn (Methodology
and History in Anthropology).
Godelier, M., T. R. Trautmann and F. E. Tjon Sie Fat (eds.). Transformations of Kinship.
Washington and London, Smithsonian Institution Press, 1998.
Godelier, M, T. Trautmann and F. E. Tjon Sie Fat. 1998. “Introduction.” In, Godelier, Trautmann and
Tjon Sie Fat (eds.), Transformations of Kinship, Washington and London, Smithsonian Institution
Press, pp. 1-58.
Harary, Frank. 1953. On the notion of balance of a signed graph. Michigan Mathematical Journal
2(2): 143-146.
Houseman, Michael and Douglas R. White. 1998a. Taking Sides: Marriage Networks and Dravidian
Kinship in Lowland South America. In Transformations of Kinship. pp. 214-243, in eds. Maurice
Godelier, Thomas Trautmann and F. Tjon Sie Fat. Washington, DC: Smithsonian Institution Press.
http://eclectic.ss.uci.edu/~drwhite/pub/SIDES5.pdf
Houseman, Michael and Douglas R. White. 1998b. Network Mediation of Exchange Structures:
Ambilateral Sidedness and Property Flows in Pul Eliya. Kinship, Networks and Exchange,
Chapter 4, pp. 58-88. Edited by Thomas Schweizer and Douglas R. White. Cambridge:
Cambridge University Press. pp. 59-93. http://eclectic.ss.uci.edu/~drwhite/pub/PULCAMB1a.pdf
Kolenda, Pauline. 2010. Personal communication in reference to Nattathi Nadars of southernmost
Tamilnadu.
Leaf, Murray. 2010. Personal communication.
14
Lehman, F. K. 2010. Personal communication.
Lévi-Strauss, Claude. 1969. The Elementary Structures of Kinship. Boston: Beacon Press.
Lounsbury. Floyd. 1964. The Structural Analysis of Kinship Semantics. Proceedings of the Ninth
International Congress of Linguists, ed. Horace G. Lunt, pp 1073-1093.
Morgan, Lewis Henry. 1981. Systems of consanguinity and affinity of the human family.
Washington, DC: Smithsonian Institution.
Read, D. 2001. What is Kinship? In The Cultural Analysis of Kinship: The Legacy of David
Schneider and Its Implications for Anthropological Relativism, R. Feinberg and M. Ottenheimer
eds. University of Illinois Press, Urbana. Pp. 78-117.
Read, D. 2007. Kinship theory: A paradigm shift. Ethnology 46:329-364.
Read, D. 2008. Formal Models and Explanatory Arguments. Structure and Dynamics:
Anthropological and Related Sciences 3(2): 123-140.
Read, D. 2009. Another Look at Kinship: Reasons Why a Paradigm Shift is Needed. Algebra
Rodtsva 12:42-69.
Read, D. 2010a. The Logic and Structure of Kinship Terminologies: Implications for Theory and
Historical Reconstructions. In Per Hage and the Renaissance in Kinship Studies. D. Jones and B.
Milkic, editors. Salt Lake City: University of Utah Press. Read, D. 1984. An algebraic account of
the American kinship terminology. Current Anthropology 25: 417-440.
Read, D. 2010b. The Generative Logic of Dravidian Language Terminologies. Submitted to
Mathematical Anthropology and Cultural Theory.
Read, D. 2010c. June 10, 2010 personal communication.
Read, D. and C. Behrens. 1990. KAES: An expert system for the algebraic analysis of kinship
terminologies. J. of Quantitative Anthropology 2:353-393.
Tjon Sie Fat, F. E., and T. R. Trautmann. 1998. On the Formal Analysis of “Dravidian,” “Iroquois,”
and “Generational” Varieties as Nearly Associative Combinations. Pp. 59-93, In, Godelier,
Trautmann and Tjon Sie Fat (eds.), Transformations of Kinship, Washington and London,
Smithsonian Institution Press
Trautmann, T. R. 1981. Dravidian kinship. Cambridge: Cambridge University Press, 1981.
White, Douglas R. 1999. Controlled Simulation of Marriage Systems. Journal of Artificial Societies
and Social Simulation 2(3): article 5. http://jasss.soc.surrey.ac.uk/2/3/5.html
White Douglas R., and Ulla C. Johansen. 2005. Network Analysis and Ethnographic Problems:
Process Models of a Turkish Nomad Clan. New York: Lexington Press.
White, Douglas R., and Paul Jorion. 1992. Representing and Computing Kinship: A New Approach.
Current Anthropology 33(4): 454-463. http://eclectic.ss.uci.edu/~drwhite/pw/White-Jorion1992.pdf
15